CT图像肺结节的全自动算法研究

Automated algorithm for CT image of pulmonary nodules

肺癌是现在发病率和死亡率最高的癌症,其对人类的健康和生命有着巨大的威胁。对肺癌的及时诊断可以有效提高患者的生存率,肺肿瘤早期在CT图像上表现为肺结节,因此,对CT图像中肺结节的研究在辅助肺癌诊疗方面有着重要意义。针对肺结节的识别和分割,提出了一种基于卷积神经网络(CNN)和改进随机游走(RW)的新算法。首先,通过结合CNN实现了对肺结节的全自动识别和检测,其次,通过改进随机游走算法的权函数,结合灰度特征和纹理特征,提高了对肺结节的分割效果。通过实验对比医生手动分割结果(金标准),本文方法对肺结节区域的分割结果的Jaccard系数平均值在0. 75以上;面积重合率平均值为0. 858;单幅图像的平均分割时间均不超过6 s,证明了本文方法具有全自动化,分割速度快,精度较高,鲁棒性好等特点,或可有效辅助临床医生对肺癌的诊疗。

Lung cancer is the cancer with the highest morbidity and mortality, which poses a huge threat to human health and life. The timely diagnosis of lung cancer can effectively improve the survival rate of patients. The early stage of lung cancer is pulmonary nodules on CT images. Therefore, the study of pulmonary nodules in CT images is of great significance in the diagnosis and treatment of lung cancer. In this paper, a new algorithm based on convolutional neural network ( CNN) and improved random walk ( RW) is proposed for the identification and segmentation of pulmonary nodules. Firstly, Automatic identification and detection of pulmonary nodules by combining CNN. Secondly, by improving the weight function of the random walk algorithm, combined with the gray features and texture features, the segmentation effect on the lung nodules is improved. By comparing the doctor’s manual segmentation results ( gold standard ), the Jaccard coefficient of the segmentation results of the lung nodule region is above 0. 75;the average area coincidence rate is 0. 858;the average split time of a single image is no more than 6 seconds. It proves that the method is fully automated, with fast segmentation speed, high precision and good robustness, which can effectively assist clinicians in the diagnosis and treatment of lung cancer.